Re: Rationale for /etc/init.d/* being conffiles?
sanvila@unex.es (Santiago Vila) wrote on 20.12.97 in <[🔎] Pine.LNX.3.96.971220174627.1163A-100000@cantor.unex.es>:
> Could somebody please explain the rationale for having *all*
> /etc/init.d/* scripts as conffiles?
And you were given that rationale. Repeatedly.
> Date: Fri, 19 Dec 1997 15:23:49 +0100
> From: Marco d'Itri <md@linux.it>
>
> I find useful to modify some of the scripts. (e.g. I don't need RPC and
> I use different command line options for sendmail.)
>
> I'm not saying that *any* of them should be a conffile. I want to know why
> *all* of them should be *by policy*. You gave me examples of some scripts
> that you find useful to modify. Ok, let's make them conffiles. But should
> they be conffiles because they are useful to modify or because policy says
> so?
Policy says so because they are useful to modify. What is so hard to
understand about this?
I'm beginning to feel like being trolled.
> This does still not explain why the policy mandates *all* scripts to be
> conffiles.
Yes, it does.
> Yes, you can change the behaviour of the program by modifying the script.
> But this is also true for every script in /usr/bin. Why don't make
> conffiles all scripts in /usr/bin by policy also?
1. Because people want to do this _FAR_ less often.
2. Because MD5 checking isn't cheap.
Really. #2 is the only good reason why we don't have every file being a
conffile.
3. Because "everything below /etc is a conffile" is what people expect.
After all, that's the FSSTND definition of /etc.
> Ok, let's do it for init.d script which are likely to be changed.
That's all of them.
> But this does not explain why all of them should be conffiles by policy.
Yes, it does.
> Date: 19 Dec 1997 21:11:00 +0200
> From: Kai Henningsen <kaih@khms.westfalen.de>
>
> > Could somebody please explain the rationale for having *all*
> > /etc/init.d/* scripts as conffiles?
>
> Because they are.
>
> Nice answer! This is a chicken and egg problem...
Not at all. There's absolutely no circular dependency here.
> This is just another example that shows that /etc/init.d/diald should be a
> conffile, but examples are just examples. Why a script that is useful to
> be modified should imply that *all* scripts are useful to be modified?
I cannot believe that I really need to explain this to you. It should be
painfully obvious.
For me, the modified scripts are watchdog, diald, xntp3, and xdm. For the
next guy, they'll be different. Nearly all of them will be changed by
someone.
What's the use of searching for the one or two scripts that aren't changed
by anybody - just to find that the next guy to install Debian _will_
change them?
> Some of them are useful to be modified, some of them not. Let's make
> conffiles those that are useful to be modified and not those that not.
I have yet to hear a convincing argument that there _are_ "those that
not".
MfG Kai
Reply to: